train_results.json

{
    "epoch": 0.96,
    "num_input_tokens_seen": 18874368,
    "total_flos": 8.352389743383675e+17,
    "train_loss": 0.9142353534698486,
    "train_runtime": 994.2403,
    "train_samples_per_second": 4.782,
    "train_steps_per_second": 0.009
}